It is indeed the CMS FORMAT that takes very long. The idea of Dennis to format separataly is very good, but instead of generating a small pool first following by FILEPOOL MINIDISK, one could FORMAT&RESERVE all the SFS minidisks in batch workers, remove the FORMAT option from the DMSPARMS file and then use FILESERV GENERATE to directly create the pool at full size.
To gain more time: use DFSMS COPY to format the target minidisks: one copies a 1 cylinder T-disk onto the target, followed by a FORMAT RECOMP: 'CP DEFINE T3390 1 1' queue 'YES'; queue 'TEMP' 'FORMAT 1 Z' 'DFSMS COPY Z 1191' arg(1) 'ACCESS 1191 T (RECOMP' On a z9 with DS8000 disks, DFSMS with the default OPTIMIZE 3 costs 9 seconds as opposed to 12 seconds for a 500 cylinder minidisk. OPTIMIZE 4 reduces this to 5 seconds. R V.cpu T.cpu SIO Elaps Page Page ! Comment/Command U time time time rds wrts ! N (Secs) (Secs) nbr (Secs) ! ------------------------------------------+----------------- 1 0.01470 0.14401 7.5K 26.11 0 0 format 1191 T 2 0.05736 0.14871 1.5K 8.769 0 0 EXEC tt 3 0.11595 0.21907 532 5.262 0 0 EXEC tt (OPT 4 Ready KRIS at VMKBNL01 ; T=0.12/0.22 10:12:06 DFSMS is a free feature of VM, but is not preinstalled. 2008/3/30, O'Brien, Dennis L <Dennis.L.O'[EMAIL PROTECTED]>: > Marcy, > FILESERV GENERATE does a CMS FORMAT of all the server minidisks, one at > a time. If you want to speed it up, you could do the FILESERV GENERATE > with a minimum number of minidisks in your POOLDEF file. While that's > running, run several VM:Batch jobs in parallel to CMS FORMAT and RESERVE > all of the other minidisks. When everything is formatted, run the > FILEPOOL MINIDISK command to add the rest of the minidisks to the > server. > > Dennis O'Brien > > "A society that gets rid of all its troublemakers goes downhill." -- > Robert A. Heinlein > > > > -----Original Message----- > From: The IBM z/VM Operating System [mailto:[EMAIL PROTECTED] On > > Behalf Of Marcy Cortes > Sent: Saturday, March 29, 2008 20:06 > To: [email protected] > > Subject: Re: [IBMVM] DR refresh of active SFS > > >Part of a DR procedure is running an exec that regenerates all SFS > servers > > How long does yours take? > Takes an annoyingly long time - nearly an hour - to regenerate an 90G > filepool. > If we didn't do disaster tests monthly, it wouldn't be so bad, but what > the heck could it be doing for an hour?! (z9 with ds8000 -- not > lightweight HW either). > > > > > Marcy Cortes > -- Kris Buelens, IBM Belgium, VM customer support
